Hospital data may be more accurate than census data in estimating the ethnic composition of general practice populations.

Background: Equity of service provision by age, ethnicity and sex is a key aim of Government policy in the UK. The prevalence, natural history and management of common chronic conditions, such as diabetes and hypertension, vary between ethnic groups. Developing and monitoring responsive local services requires accurate measures of ethnicity and language needs. Hence establishing the ethnic composition of GP populations is important.

Objective: To compare three methods of estimating the ethnic composition of GP registered populations in three east London primary care trusts (PCTs).

Design: Self-reported ethnicity, routinely collected at practice level (and considered the 'gold standard'), was compared with two indirect methods of attributing ethnicity. The indirect method currently used in the UK assigns ethnicity to GP populations based on geographical postcode attribution from the national census. A proposed alternative indirect method uses the ethnic breakdown of hospital admission data from practice lists to attribute ethnicity to the whole practice population. Comparisons were made between practice self-report recording and these two indirect methods. Bland-Altman plots were used to assess the agreement between methods of measurement.

Results: Data from 103 practices, covering 70% of the GP registered population, was used. The hospital admission method showed better agreement with practice self-report data than the census attributed method. For white populations Bland-Altman plots showed a mean difference of 1.4% (95% CI-14.9 to 17.7) between hospital admission and practice data, and a mean difference of 12.5% (95% CI-6.2 to 31.1) between census attributed and practice data. Differences were also found for south Asian and black populations.

Conclusion: Practice ethnicity measured using hospital attendance data is in closer agreement with practice recording of self-reported ethnicity than the census attribution method. Census attribution may provide misleading information on the ethnic composition of practice populations. We recommend that healthcare commissioners change to this method of measurement when practice self-report data is not available.